Feature #20182
Updated by Tom Clegg about 1 year ago
Currently, if many workflow runners are submitted at once, they might use up all available capacity, and they will all submit child containers and wait forever for them to run, because all available capacity is consumed.
The limit should apply to any container with the (new) scheduling parameter @supervisor: true@.
Arvados-cwl-runner should set @supervisor: true@ when submitting a supervisor container.